Checkr’s Daniel Yanisse on tackling bias in people and AI
AI can be used to reduce human bias, but if we’re not careful, it can also learn it. In this episode of Spotlight On: AI, Daniel Yanisse, Co-Founder and CEO of Checkr and Accel partner Rich Wong discuss their concerns, hopes, and advice for building a fairer future.
“Be self-aware. You know where bias may be, so be cautious where there is AI in those business processes. What becomes dangerous is when it is a black box, and you don’t think about how to apply it tactically.” – Daniel Yanisse
Born in France, Daniel's robotics passion led him from college in Switzerland to the Silicon Valley. After roles at NASA and Cisco, he was drawn to the allure of startups. In 2013, he and his co-founder, Jonathan Perichon, realized background checks were a significant bottleneck to the on-demand ecosystem emerging at the time. To fix it, they decided to build an API for background checks.
After YCombinator (S14), Checkr was born. Since Accel’s partnership in 2014, we've seen Checkr evolve into the leading background check platform with a clear mission to provide fair chance hiring opportunities to the 80 million Americans with criminal records, a mission made possible through artificial intelligence.
Creating a fair and accurate background check involves using AI to navigate a complex and messy sea of unorganized data. Checkr’s AI initially focused on crime classification and expanded to address broader hiring process issues and mitigate human prejudices. Today, it can identify human biases in the recruiting process, pinpoint discrepancies in compensation, and scrutinize interview records for potential race or gender biases.
There are many reasons to be optimistic about AI – productivity, accessibility, and reduction in human error. At the same time, there are valid concerns around issues like bias. It is a complex problem that demands a dual solution – AI can help humans by pointing out their existing biases, and humans can reciprocate the effort by training AI against it. We discuss the concept and more on this episode of Spotlight On.
- 00:00 - Daniel’s upbringing in France and early interests that inspired his move to the United States
- 05:00 - The experience with background checks that inspired the creation of Checkr
- 07:30 - Getting into YCombinator and meeting early investors through “speed-dating”
- 09:10 - Checkr’s early days meeting needs of the booming on-demand economy, and gaining early customer traction
- 11:40 - The inspiration behind Checkr’s mission to build a fairer future by reducing the imbalances and biases in the hiring process
- 19:50 – How Checkr has been using artificial intelligence since its earliest days to make sense of the background check data, ensure accurate classification of crimes, check for bias in the process
- 26:00 - Advice for founders who are excited about the potential of artificial intelligence but want to reduce AI biases that exist in machine learning models
- 35:00 - Closing thoughts and a warning about the technological divides that may come from AI progress
Host: Rich Wong, Partner at Accel
Explore more episodes from the season:
- Episode 01: AssemblyAI's Dylan Fox on building an AI company during a period of radical change
- Episode 02: Roblox’s Daniel Sturman on Building Great Teams in the AI Era
- Episode 03: Ada’s Mike Murchison on how AI is revolutionizing customer service
- Episode 04: Merge’s Shensi Ding on powering the next generation of AI SaaS companies
- Episode 05: Scale AI’s Alexandr Wang on the most powerful technological advancement of our time
- Episode 06: Bard’s Jack Krawczyk on the birth of Google’s AI chatbot and the creative potential that lies ahead
- Episode 07: Synthesia’s Victor Riparbelli on creating an environment to harness AI benefits and reduce harms
- Episode 08: Ironclad’s Cai GoGwilt on a decade of anticipating the transformative power of AI
- Episode 09: Checkr’s Daniel Yanisse on tackling bias in people and AI
- Episode 10: Cinder’s Glen Wise on trust and safety threats and holding AI accountable
- Episode 11: Transcend’s Kate Parker on putting data back into the hands of users in an AI-driven world
- Episode 12: Arm’s Rene Haas on building the brain of artificial intelligence
Rich Wong (00:13):
Daniel, welcome to Spotlight On. We are very honored to have you. This is a program where we talk about the technologies that are shaping our industry and even more importantly, the great founders and people that are shaping it. So we're really, really excited to have you.
Daniel Yanisse (00:27):
Thanks, Rich. I'm super excited to be here and to have this conversation with you.
Rich Wong (00:31):
So we've known each other a little while now. It's kind of amazing how one of the real joys of being in this business for me is getting a chance to work with founders like yourself for I think what will be decades of a business relationship and a personal friendship. And I really appreciate that with you and with Jonathan.
Daniel Yanisse (00:49):
Yeah, me too. I mean, we're going on nine years now, right? Since we first met in 2014. YC demo day.
Daniel’s background, immigrating to America and working at NASA
Rich Wong (00:56):
It seems like just a couple of years. But yeah, I think it has been nine years in all the different stages as we were talking before about your life as well. Yeah. Well, why don't we just back all the way up. I think one of the amazing things about Silicon Valley and about the startup ecosystem is where people come from. And I know you have an incredible personal story, so just describe to me a little bit about your family. Where did you grow up? What were your influences and the journey of how you ended up here in California?
Daniel Yanisse (01:24):
Yeah, yeah. So I grew up in France, born and raised in France. You can hear through my accent. My parents are immigrants, so they actually came in the eighties to France from Romania and Syria, and they're a big influence for me. I said they're my inspiration. They were doctors, but they had to work extremely hard to make it in France and integrate and learn French.
Rich Wong (01:46):
How old were they when they came from France?
Daniel Yanisse (01:49):
They were in their thirties. The thirties. So they were in the thirties. They were still had to study med school twice because the degree would not be accepted, accepted in France. So they're in their thirties, even early forties. And I would see them as a kid still passing exams and working really hard in addition to being doctor assistants and interns. So they're my motivation and inspiration. So that was my childhood in France. Really grateful to my parents and my sister and I have really grown up in France with good public schools and a great system. So that's where we started. And then I went to college in Switzerland, not very far on the French part of Switzerland. Always wanted to be an engineer, always liked building computers, tinkering, I like math, physics, things like that. And then, yeah, I was a big fan of robotics. I wanted to go and build robots and get into robotics, and I've done some of that in college and around college.
Rich Wong (02:48):
How did you end up in the United States?
Daniel Yanisse (02:51):
So when I was in Switzerland, Switzerland's a bit more international and in robotics, like the innovators, I was studying robotics, like silicon technology. So for me, the dream was more in the US, maybe companies like Intel. At some point I was like, oh, I wanted to work at Intel. So I was a bit more focused on the hardware side and most of the leading companies were American companies and tech companies. And then I came to Silicon Valley in 2010, I got an internship at NASA Ames working in the robotics lab. And so that's when I landed in the Bay
Rich Wong (03:25):
Was this the Mars Rover project.
Daniel Yanisse (03:27):
Like Mars Rovers prototypes and AI navigation algorithms. I worked on that and that's kind of where I switched from hardware to software. I realized the future is software, everything, hardware, kind of the older companies and software was eating the world, including robotics. Most of the innovation was through the software and through early AI.
Getting involved with the startup world
Rich Wong (03:48):
So you were a NASA researcher working on robotics, this Mars Rover project. When did you decide to jump out and try your hand at the startup world? And you went to a company, I believe before starting Checkr that helped create the inspiration for Checkr after leaving nasa? Yeah.
Daniel Yanisse (04:06):
Yeah. So NASA was great, but NASA is more almost like academia. It's quite slow. You work on space projects for 10 plus years, only one in 10 projects ever makes it to space. You have to be really passionate and patient, which I was not. I like doing new things and learning things, and then I discovered the whole startup world. And so I was like, wow, you can be an engineer and jump from healthcare to software for businesses to automotive. You can work in any industry.
Rich Wong (04:35):
Did that feel risky to you to leave this very safe, I'm sure very fascinating job, but as you say, probably slower at NASA than the startup world, or you didn't think about it, you taken the immigrant risk before your parents had taken the immigrant risk.
Daniel Yanisse (04:51):
So it wasn't that big of a deal. I was an unpaid researcher, so it felt safe actually to get a real job and be paid for my work. And I dunno, I was like 20 something.
Rich Wong (05:03):
Daniel Yanisse (05:04):
No attachments. I was excited by the adventure, and so I joined the startup in LA first. That's where I met my co-founder, Johnathan. We became best friends. That startup was early and failed. And then I moved to another startup in the Bay Area, and that was the leave, the last work job I had before we started Checkr, and that's where we had the idea for Checkr.
Founding Checkr and identifying a core problem to solve
Rich Wong (05:28):
Tell me a little bit about the founding idea for Checkr. How did you and Jonathan come up with the idea and tell us about those early days of how it got going.
Daniel Yanisse (05:37):
So when we are engineers, Jonathan and I, my co-founder in the same company, so we were working together at Deliv. It was a on demand delivery company. We are building all parts of the software. We're building the driver mobile app for drivers to get orders and go pick them up and deliver packages to customers.
Rich Wong (05:58):
Sort of like an Instacart.
Daniel Yanisse (05:59):
Early Instacart, DoorDash with the very early days of the on-demand economy. And so as part of this app, we're struggling to hire enough drivers. And one of the bottlenecks in the hiring process for drivers was the background check step. It was taking long, it was manual, it was outside the system. So my CEO at the time tasked me to go find a better background check vendor, something maybe with an API we could integrate to streamline the onboarding flow for drivers.
Rich Wong (06:30):
How did you find those partners or suppliers at the time? How many of them even had an API in the way we think about it ?
Daniel Yanisse (06:36):
So I just went on the internet. I googled background checks, background checks, APIs. I found a few vendors, but I was surprised that they were not really software companies. They were very old school manual process companies and they didn't have APIs or they had a very old API that was confidential. I was outraged as a developer. It was 2014. They were things like Stripe and AWS and Google Maps. They were open APIs everywhere.
Rich Wong (07:05):
So the existing incumbents just didn't feel modern to you. They just were old.
Daniel Yanisse (07:12):
They were old, more like enterprise and did not even want to work with our startup back then. They were like, you're too much of a startup. We work with big enterprise companies. Maybe in six months if you're bigger and wait in line, we'll give you access to implement you as an API customer. So that was frustrating. And so I said to Jonathan, we can build something better than that.
Daniel’s Y Combinator experience and pitching his first wave of customers
Rich Wong (07:33):
And you went through Y Combinator, which was a great experience, as you've said to me multiple times, right?
Daniel Yanisse (07:39):
Yeah, it was really great. I was already a Y Combinator fan, even back in Europe. I was reading everything, and I was like, that's a great route. You can go from being an engineer to being a founder. That's the dream. And so big fan of Y Combinator. I went to different events and eventually we applied and we got in with this idea.
Rich Wong (08:00):
We actually met for the first time at a traditional demo day. And I remember watching your and Jonathan's presentation and being so impressed, but we had not ever met or had any interaction prior to that, and now it's been nine years.
Daniel Yanisse (08:14):
I like to tell people that the way we met was almost like speed dating. It was like speed dating with hundreds of investors in that big room in the computer museum. And we had a lot of interest. A lot of different investors would come to me and there was even I think a Chinese investor who came and said, I will write you a $20,000 check if you shake my hand now in the room. And I was like, I dunno. And I knew nothing about the Brazil landscape and who are the good investors. And so I always tell people that I got really lucky in that speed dating to meet you, and it's been a fantastic partnership.
Rich Wong (08:49):
So Daniel, let's just talk about the early days of how you got started in your first customers and what segment you focused on.
Daniel Yanisse (08:56):
Yes, so we built the product that we needed in our previous job, so really automated background check API to streamline the hiring of delivery drivers. And so we're lucky that in 2014 it was the early days of the on-demand economy. Lots of startups were funded. It was like Uber for X in every category. Uber itself was very early. It was only doing Uber Black back then. The black cars only, no ride sharing, but it was a big category and a lot of those startups were funded in Y Combinator. So one thing I love about Y Combinator, it's you can find your early customers in that network.
Rich Wong (09:36):
So Instacart, DoorDash, these companies existed, but they were small.
Daniel Yanisse (09:40):
Yeah, they existed and they launched I think one year before we launched. So it shows you they were in their first year, but raising a lot of money and growing really quickly. And so we got them as early customers, which really helped us get product market fit.
Rich Wong (09:53):
How did you convince these founders that this is a pretty, as we agree now, a very important security and trust and safety step to keeping your service safe? How did you convince Tony at DoorDash or the Uber folks to trust a small startup like Checkr was in those early days?
Daniel Yanisse (10:12):
Yeah, so the good news is they were also small startups, A little bit bigger but not much more. And it was quite easy. I was really lucky, literally, I would just email people and say, we're building a background check API, we know it's painful. And literally people would say, oh my God, that's the biggest problem I have. Please come. And we go to the meeting and they literally say, welcome. We hate our current solution. What you guys are building sounds like exactly what we need. It was quite unreal. It's not usually how it happens now when I'm launching new products, we have to hustle much more. But there was really an acute problem for most customers would just welcome us and want to try the product asap.
Rich Wong (10:53):
I think as you described the story of how you came upon the problem and how you built the company around a problem that you had actually tried to work on yourself when you were the customer. I think that that experience, that intuition makes it so much more likely to succeed than one tries to dream up the idea off a clean sheet of paper. You are solving a problem that you felt yourself.
Daniel Yanisse (11:15):
I see that often, especially for B2B companies, oftentimes the founders have some kind of unfair knowledge about the problem and the experience. So that's how they know what first problem to build deeply because they are the customer. So I've seen that often it's a good pattern for B2B businesses.
Checkr’s evolving mission, and building a fairer future
Rich Wong (11:36):
So let's talk a bit about the mission of Checkr. And I know you thought very deeply about this with Jonathan, what Checkr stands for. Why don't you just talk about what is the vision and the mission of what you want Checkr to achieve? And then we'll talk about some of the things that come after that.
Daniel Yanisse (11:49):
Yes, yes. So when we started the company, we didn't have a mission, our mission to solve customer problems, to build a great product for customers,
Rich Wong (11:58):
A good starting mission is to succeed. And so to solve real problems.
Daniel Yanisse (12:01):
We wanted to help customers and have a product that's useful. So we started that. But as we were doing the first year of background checks and started to process background checks, talk to consumers and job candidates who are going through the process, talking to customers and seeing how they're making hiring decisions, we realized a few things. So first, of course, background checks. Everyone and all customers were convinced about the value of safety, the value of understanding people's background, building trust through that data. It was a standard part of the hiring process, but it was very skewed towards finding the bad guys and rejecting them. That was the only thinking about background checks. And the whole industry was pretty focused on fear, like the marketing tactics where they might be a criminal in your business, you have to find them otherwise you are at huge liability risk and safety risk, which is really exaggerated in a way.
(12:56): And we realized that it's not a black and white. There's actually thousands of job applicants who yes, don't have a perfect background, but extremely motivated and ready to work. And so we felt already in the very days there's a big imbalance in the process and that leaves thousands of people on the side of the road who many of them have a minor crime of a crime or mistake. That was a long time ago. That's not relevant to the job anymore. But the way the background check process is set up is as a customer, if you get two candidates and one has a long list of flags and the other one is like nothing, you're just going to do the pragmatic decision of moving on with a simpler candidate. And so the product was also not set up to show you this spectrum and balance. So quite early on we said our mission should be to be different from the rest of the industry and make this industry better because we believe with technology we can improve fairness and we can change this process to not be binary and help businesses decide who are the best candidates for them.
(14:02): And we already knew that there's going to be lots more people who are actually being rejected who could be a great fit for businesses, and we could change that balance over time. Accellent. So that's how we set our mission to build a fair future, which will help through a better product to give opportunities to millions of people, to get great jobs.
Rich Wong (14:21):
To build a fairer future.
Daniel Yanisse (14:23):
To build a fairer future. And by changing some of those decisions and helping customers see through the different cases and data actually that because we have a lot of scale that enables millions of people who are rejected to now get a shot at opportunities, which is really fantastic.
Rich Wong (14:43):
It is fantastic. How many people are we talking about that have something in their backgrounds that could otherwise prevent them from getting a job.
Daniel Yanisse (14:49):
So now if we look last year, we do a check, we do tens of millions of background checks a year. So we're getting a pretty big scale. And last year alone, there's 2 million people who had some kind of flag that went through our system, but at the end of the day, the employer decided to move forward.
Rich Wong (15:09):
2 million people.
Daniel Yanisse (15:10):
2 million people.
Rich Wong (15:11):
And what do you think it is in the overall US population? How many people in the overall US workforce have some sort of flag or potential flag in the background?
Daniel Yanisse (15:21):
It's a lot, right? The flag can be as simple as speeding violation. Sometimes a D-U-I minor infraction like possession of marijuana. Those are very high volume in the us. There's one in three Americans has some kind of flag of violation.
Rich Wong (15:41):
One in three Americans, right? So if you don't help create a system where people understand how to treat that fairly, one in three Americans could be excluded or at risk.
Daniel Yanisse (15:55):
Are being excluded. And sometimes when that is not a problem at all for the business, but you have to put yourself in the shoes of the recruiters and the HR departments who are getting thousands of resumes and need to make decisions fast. So that's where we come in and we try to help them create rules and criteria that's tailored for their business that allows them to streamline some of those decisions.
Fair chance hiring, and practicing what you preach
Rich Wong (16:19):
To this point about fairness and the mission of Checkr, you have been a huge supporter and really evangelist as a company for the concept of fair chance hiring. Can you just talk a little about what is Fair Chance hiring and what are we actually doing to support that cause?
Daniel Yanisse (16:35):
Yeah, so fair chance hiring also sometimes called Second Chance hiring is giving second chances to people. So it's basically helping someone who has a criminal record and interviewing them, building trusts, and then giving them a job, giving them a chance. And so we've been a huge fan of that. We've partnered with many nonprofits. I myself went to prison educational systems and programs to better understand how people are getting trained and released from prison. And I met fantastic people. I met some of the best, most motivated, warmest people I've ever met. And it opened my eyes. I was like, wow, those people, I'd love to hire them in my company. And so we did. We started to hire some of the best of the best.
Daniel Yanisse (17:23):
In Checkr itself, yes, we started to hire a few people in some of our teams to onboard them and train them to work in tech. And they've been some of our highest performers because if you give someone a chance and who's been rejected hundreds and hundreds of time, and their goal going out of prison is to turn their life around to prove to the entire world to their family that they can be a better person than what they've known for in the past. And so if you align that self-motivation with a job in tech, that can create some of the best win-win situation and most motivated employees, I think it's very counterintuitive. There's lots of bias and fear around it, but when you actually meet and interview the people, it doesn't mean hiring everyone with a record. It means giving people a chance to interview and to learn their story and do they own their mistake and are they ready to move forward? And many people are. And so then it's been a success for us at Checkr and now we try to bring that success to other businesses, to our customers because there's still a huge shortage of talent in so many industries. And again, it's one in three Americans. So it's good for the economy and for the job markets to explore these opportunities and we as a company help companies and businesses benefit from that.
How Checkr uses AI and machine learning to wrangle disorganized data
Rich Wong (18:39):
So Daniel, when you first built the Checkr product, part of it was a modern API as opposed to sort of the manual sort of stodgier companies that were in the industry before, but you also thought a lot about building machine learning and AI into the system. Talk a little bit about where you started on the product as it relates to AI and then we'll talk about the evolution to a lot of the discussion in the market today.
Daniel Yanisse (19:02):
Yes, yes. So I've always been an AI fan. Even at NASA, I was working on computer vision algorithm and navigation, autonomous navigation on Mars. So I love AI, but I'm also not just the technologies. At the end of the day, we build products to solve customer problems. And sometimes you don't need fancy technology to solve a customer problem. The customer just wants an email button, like simple things. But for us in background checks, the big problem is to build an accurate background check. You have to collect a lot of data. It's actually very complex and the data is very messy and unorganized. So in order to make sense of the data so that customers can actually read and understand the background checks, we had to use AI to classify and organize the data. For example, in the us, every single county and state has different language for the different crimes.
(19:59): So theft is not called the same in different states and it has codes and everything. So to organize this data, we built NLP classifiers. So we basically took millions of records and we built categories so we can say this is A DUI or this is theft or this is possession of a gun. And so we created those categories and that allowed us to then create a product where customers can say, I don't care about possession of marijuana. But that was not possible in the past because it was just free text that you couldn't really organize. So that was our first application of AI to make sense of the background check data.
Rich Wong (20:37):
So technically it was a classifier classifier, it would organize and classify information that otherwise would be extremely hard to categorize by.
Daniel Yanisse (20:47):
Yes, exactly. So we used AI, more traditional AI back in the days before genAI was invented to classification of the background check data. Also, another problem have is to do the name matching to make sure that this record with a few partial piece of information is belonging to that person. Very important of course. And the accuracy in the industry was quite low. So we also used AI to have the best accuracy on name matching to records. So there are two simple use cases. You can use AI for many, many problems, but those are two early use cases where we built AI to improve the accuracy and the quality of background check.
Rich Wong (21:24):
The greatest area of focus these days, at least in the press, in the tech press, is around generative AI. But the application of how Checkr built its first product was really about the classification and matching and the ability to synthesize across millions of records to find meaning not necessarily on the generative basis.
Opportunities and concerns with generative AI in hiring, and addressing the problem of bias
Daniel Yanisse (21:44):
But I think the world has really changed last year since ChatGPT and engineer algorithms, which I think is exciting. If we look to our customers and our products, our customers are mostly in hr, in recruiting, in operations teams, the whole recruiting process and background check process is regulated, accuracy and compliance extremely important. So I would say before there were a lot of fears in recruiting technology and HR tech around AI. I think everyone's in those articles of like, is AI bringing bias into the hiring process by automating decisions away from humans without knowing how it's done? Is that going to lead to more discrimination, which I think are valid concerns definitely for the industry. But I do think since Gen AI happened, there's been some change. It's so powerful and every business, every department has to at least explore how this technology can improve the business.
(22:47): And so with customers, I'm seeing more openness to AI, but we have to be careful of what are the applications, right? But you see, you can use the AI not to make the hiring decision for you, but to, for example, better understand information about the candidates. You can even ask the ai, Hey, find bias in my recruiting process. Oh, interesting. You can ask AI anything, right? You could say, Hey, help me improve, find bias, find discrepancies in compensation across employees in the company, or look at the interview records and tell me if there's bias of gender or of race in some of the interviews happening.
Rich Wong (23:28):
So if I play that back to you, right, in many ways you think the AI can be used to ask questions of your own data and try to find both the good things, but even maybe some of the biases that might be embedded in even our hiring processes at. Is that roughly right?
Daniel Yanisse (23:46):
Yeah, I think AI can be used for anything you want it to. And so there are some dangerous areas and in recruiting, we have to be very mindful of is there bias in the process? Where is the data coming from? How has it been trained? And there is bias in the internet and GenAI has been trained on the entire internet, and it's just this human bias because humans are biased. So I think we're going to have to be more careful in hiring and compliance areas on how we use AI. We're going to have to really test and confirm that it's working. It's not doing unintended things. Quality reviews are going to be very important, but I think there are lots of exciting applications for us. For example, we are doing two things in AI in our company. So first we enable the whole company to play with AI and see and tinker and see what's possible in marketing and finance, in product, in engineering to improve productivity. We did that by, I think there's a lot of fear in companies around privacy. Is your IP protected? Is your customer information not going out to the AI algorithms? So we put in place different layers of protection so that our sensitive data does not go out, and those tools can be used safely within the company. But then we really enabled, and I want to give kudos to our CFO Naeem who's been really leading that task.
Rich Wong (25:08):
Daniel Yanisse (25:09):
Naeem is our CFO, and he's huge on AI and he enabled the whole company to really play with AI in the very early days. And it's been great because otherwise people can't see the innovation and the opportunities, and we have a lot of great use cases. Our marketing team is leveraging extensively to be more productive. We're testing copilot with our engineers. There's lots of great internal applications already.
Using AI to enhance productivity, and Daniel’s advice to CEOs on identifying and managing bias
Rich Wong (25:34):
What do you think are the most likely candidates to help Checkr's productivity in terms of the use of AI for Checkr?
Daniel Yanisse (25:41):
Yeah. I mean internally there's use cases. Like I said, marketing and writing copy and helping you write better, faster has been great. We use it to review our internal business document and strategy. I use sometime to draft emails or documents and then you edit it, but it gives you a boiler plate really quickly and can review your writing, make it tighter. So everything around writing is working really well. And then I think the big opportunity, which we don't know yet how productive is going to be, but it's around engineering and helping write code, helping write tests, helping doing code reviews. If we can add 10, 20% of engineering productivity, it's incredible for huge difference a product company like us.
Rich Wong (26:25):
So you mentioned two, maybe possibly counterbalancing things. So one, you're a believer as I am, and AI is a huge help to automation, to productivity, makes companies more efficient, makes our customers more efficient. At the same time, I think you acknowledge a valid concern people might have about bias built into the AI. If you're a founder or you're a CEO of a midsize or even larger company and you obviously want the productivity benefits, but you have the questions about how to make sure you manage any bias that exists, what advice would you give that founder or CEO?
Daniel Yanisse (27:04):
Yeah, I mean, I think it's just to be self-aware and going into it eyes open where bias can and or might already be applying, especially around hiring people or making decisions on people's performance. That's where I think we have the most sensitive bias around HR and recruiting. So if there is AI around that process, I'd be very cautious and kind of A/B testing it to make sure it's not adding too much bias, but that's very narrow. I think if you then break down all of the business processes, including interview and recruiting processes, I think you can apply AI tactically in different areas. And yeah, I think what's dangerous is when it's a black box and you don't really think about it. For example, we hear a lot about using AI to scan through resumes. If you just throw 10,000 resumes and say, which one should I look at without really thinking about the criteria, that's kind of dangerous and likely to put bias in your process.
Where Checkr is growing and the power of small language models
Rich Wong (28:06):
Where would you like to see Checkr advance its mission even beyond the businesses you're in today, where would you aspire to go beyond the growth that you've already had so far?
Daniel Yanisse (28:17):
Yes. So we love solving this complex compliance, HR people problems for customers, right? Sometimes it's a bit boring compliance and it's not the sexiest field, but we actually love taking software and technology to take those hard problems and make them automated and delight for customers. And the customers love it when we do that hard work for them and it's easy. So we've done it in background checks and now we also love to do APIs and embeddable products and software. And so we are looking at the other parts of the hiring process, especially for the flexible workforce, like for gig economy, freelance remote work. We want to continue to build infrastructure products to automate some of these hard HR building blocks. So we are looking at, we launched a product called Check Onboard, which helps automate and simplify and have a great experience for the onboarding experience for job applicants. We launch a product called Checkr Pay to help 1099 workers, gig workers be paid instantly every day with a very nice experience as well. And so those are the type of new products we're launching to continue to build infrastructure for the workforce.
Rich Wong (29:31):
When you look at all of the money that's going into Anthropic or OpenAI Chat GPT or Google Bard, can you handicap for us your prediction of is this a market that will be owned, these underlying LLM models? Will it be owned by Big Tech, will open source win? It's a tough question, but what is your prediction about how this AI landscape will evolve over the next 10 years?
Daniel Yanisse (29:56):
Yeah, yeah. I mean, I'm excited about the technology. I think large language models are awesome. I think also now I'm able getting more excited about small language models.
Rich Wong (30:08):
Small language models. Say more about that.
Daniel Yanisse (30:09):
A large language model is like ChatGPT, it's trained on the entire internet. You can do everything. We also can do too much in a way, and it's inaccurate. So small language model is kind of taking the same approach, but building smaller models that are very specialized to do one task or one category of tasks. So for example, to do customer support for your company and answer all of the questions your customers, because you don't want that chat to start debating philosophy with your customer saying the wrong thing. So I think more specialized model, and I'm learning this from Databricks and other companies, I think that's very exciting technology.
Rich Wong (30:44):
So purpose-built small language models.
Navigating technological hype cycles and the potential of AI in the long-run
Daniel Yanisse (30:48):
And that makes sense. The LLMs are very expensive to run, very complex, kind of a black box. And so in business, we want to solve some specific customer problems. We don't want to solve the whole internet knowledge. And so I'm pretty excited about this technology. I think it has a lot of legs. I don't know who's going to win. I think it's overhyped right now. Like many technology cycles, there's like too much craziness. Who knows? Maybe one startup wins one big company. I can't predict. I think it's also a bit too over hyped. Even for regular businesses, you just put AI, you have to put AI in your features, and I'm trying to be pragmatic. Again, AI is a good technology as long as it solves a real customer problem. So it's a bit of a hype, which is okay. I think it's good for innovation, but I think it's good to focus back on what customers need because customers, they just want you to help them improve their life, whether it's AI or not.
Rich Wong (31:47):
To draw the analogy to AI, I know that the first startup you worked at after NASA was a mobile company. And I remember mobile, it has massive impact on our economy and the tech industry and the medium and long run. But in the short run, there were a lot of spikes of hype that didn't, the reality never quite lived up to the initial hype in some of those categories.
Daniel Yanisse (32:06):
And who in mobile, it's more like Apple and Google.
Rich Wong (32:10):
Well, until Android and iOS and the iPhone came along, there were a lot of cycles where it didn't quite live up to its potential, but in the medium and long run has had dramatic impact.
Daniel Yanisse (32:22):
Often the time, the early days, we get kind of overexcited about the potential. Then there's a few years of reality check that is not fully ready, and then over 10 plus years it becomes mainstream. So yeah, that sounds like we've seen that a bit with crypto and Web3 and many things.
Open source models and the cost problem in AI
Rich Wong (32:39):
Do you think the open source models, when we're playing around with the technology, do you think the open source models are viable competitors to these expensive but very powerful open AI or we're not sure yet? It's hard. It's so early. It's hard to know ?
Daniel Yanisse (32:57):
These, I think the cost is a problem, even though OpenAI is very expensive. I mean, it is just so big and it's because the cost of good sorts are really expensive. Those GPUs and the data centers you need are insane. So that's going to get commoditized and lower cost. So I think the cost will drive. I think right now I hear more and more companies who are rather not use the out of the box at scale because too expensive. And so maybe use open source plus other part of the stack to optimize cost. If it becomes super cheap and mainstream on AWS, on OpenAI and everything, I'm sure we'll move more to using those APIs and not run it as much in house. Right now it looks like people are going pretty deep themselves because of that cost. Cost or specialization training the algorithm on their data versus being worried about feeding the big machine.
Rich Wong (33:59):
Do you mean it's expensive for open AI or that the licenses are expensive?
Daniel Yanisse (34:03):
For the customer.
Rich Wong (34:04):
To run? For the customer?
Daniel Yanisse (34:04):
For the customer, because it sense per, it's like fraction of cent. But if you use it on millions of customers on questions, it can blow up pretty quickly. And so that's what I'm talking about, the cost to the customer, which I'm sure is going to go down quickly.
Balancing optimism and caution, and evolving AI in a thoughtful way
Rich Wong (34:22):
As we talk about this AI trend, which is a huge wave that's sweeping the tech industry and the sort of broad awareness in society about this, is there anything about that concern you or do you feel pretty comfortable that we're all evolving this technology in a thoughtful way?
Daniel Yanisse (34:40):
Technologies would always go as fast as possible to evolve the technology. There are some people thinking about, Hey, is this dangerous and risky? Are we going to have AI run the world? And it's going to become matrix and slave humans into it, but people will still push the boundary until it starts to become dangerous, actually. So I dunno, I still feel like we're far from that, right? But it is really an impressive technology that we don't know exactly how far it's going to go when we look at the development. What I'm worried about is kind like the technological divides, and I dunno if there's anything we can do about it, but technology is just moving so fast every single year, and we're fortunate to be a technologist and we keep up with everything. But if you look at everyone else in the world, it's just very hard for people to keep up, to retrain themselves into the skills you need to compete in this new world. And so that technological divide just keeps accelerating. I dunno if it's just AI or not, but AI is really going even faster. So I hope AI will help. People are also train themselves, educate, improve productivity. But yeah, I'm a bit more worried about the impact on people who can't keep up with all of these technological changes. Are you worried about some aspects of AI in the future?
Rich Wong (36:07):
I am still trying to make up my mind is the truth. I think there are a lot of reasons to be optimistic. I think you think about where we are in the American economy and even some of the demographic issues across the world where software and AI are probably the greatest chance at improving productivity, and that productivity is key of a household income per capita, which leads to a better life for people, for humans. And so I think the reason why you and I work in the technology industry is in addition to being interesting and fun, is because I think it has a major impact on how our economy moves forward and how people ultimately live a better life. On the flip side, I have some of the same questions that you just brought up around a digital divide where there are people that will be beneficiaries of this new technology and there's going to be some level of job displacement and some level of not so positive impact. And so I think things that you've really been a thought leader on fair chance hiring are attempts to try to reach across that divide. But I think we're going to need even more of those efforts as new technologies like AI become even more pervasive. So I have a lot of optimism, but frankly, I have a lot of similar concerns to what you bring up.
Daniel Yanisse (37:21):
Yeah. No, I agree, rich, and I think we will have to do our best we can from our jobs, our companies, to improve that and reduce that gap over time.
Checkr’s hopes and aspirations over the coming decade
Rich Wong (37:31):
Daniel, you've been the founder and CEO of Checkr for the last nine years. What are your aspirations for the next 10 years? What year do you want to see Checkr as we continue to grow and move forward?
Daniel Yanisse (37:42):
Yeah, no, I mean, it is been an amazing journey for me. I'm so grateful for the last nine years and the team we have, it still feels like a early startup, so I really enjoy working with customers, working with our employees, launching new products. And so for us, we're going to kind of continue to multiply ourselves as a startup, launch new startups, new products especially. We love that problem we're working on around the workforce. It's great when you can have technology that has a direct impact on jobs, on hiring, on the quality of jobs that people have. So yeah, we're excited over the next few 10 years, we will continue to build the company to be a leader in the space. We're going to bring that fairness mission that we have in all of our products, of our customer conversations. And so, yeah, so much opportunity ahead. We are really excited.
Rich Wong (38:33):
Well, Daniel, I've said this to you multiple times, but I thank you and Jonathan for letting us be a small part of your journey. It's really been a pleasure. It's just the beginning still. We've got a long ways to go, but we're honored to be one of your partners. Thank you.
Daniel Yanisse (38:46):
Thank you, Rich. It's been so fun.